English

Verifying And Interpreting Neural Networks using Finite Automata

Formal Languages and Automata Theory 2023-09-28 v3 Machine Learning

Abstract

Verifying properties and interpreting the behaviour of deep neural networks (DNN) is an important task given their ubiquitous use in applications, including safety-critical ones, and their black-box nature. We propose an automata-theoric approach to tackling problems arising in DNN analysis. We show that the input-output behaviour of a DNN can be captured precisely by a (special) weak B\"uchi automaton and we show how these can be used to address common verification and interpretation tasks of DNN like adversarial robustness or minimum sufficient reasons.

Keywords

Cite

@article{arxiv.2211.01022,
  title  = {Verifying And Interpreting Neural Networks using Finite Automata},
  author = {Marco Sälzer and Eric Alsmann and Florian Bruse and Martin Lange},
  journal= {arXiv preprint arXiv:2211.01022},
  year   = {2023}
}
R2 v1 2026-06-28T05:00:09.703Z